Operationalising Architecture’s Core Competency - Agent-based Parametric Semiology

PATRIK SCHUMACHER, Zaha Hadid Architects, London, https://orcid.org/0000-0002-2270-395X

Published in: DC  I/O 2020 , Design Computation Input/Output Conference: Algorithms, Cognitions, Cultures

1. Fig.1  Zaha Hadid Architects: Corporate space planning  and interior design for Sberbank technology centre, Moscow.  The agent-based occupancy  simulation in 1st person perspective. ZHA Agent-based Parametric Semiology (ABPS) research team: Tyson Hosmer (leader), Soungmin Yu, Sobitha Ravichandran, Ziming He.

Architecture’s social functionality resides to a large extent in its communicative capacity. The built environment orders social processes through its pattern of spatial distinctions and connections that in turn facilitates a desired pattern of social events. The functioning of the desired social interaction scenarios depends on the participants’ successful orientation and navigation within the designed environment. The built environment, with its complex matrix of territorial distinctions, is a navigable, information-rich interface of communication. To order and articulate this interface is the core competency of architecture. This core competency is reckoning with users as sentient, socialized actors who use the built environment as an orienting matrix and text of instructions within which myriads of nuanced social protocols are inscribed.

1. ALL DESIGN IS COMMUNICATION

Before a specific interaction event can commence, relevant participants must find each other, gather and configure into a constellation germane to the desired interaction scenario. Their respective expectations, moods and modes of behaviour must be mutually complementary, i.e. they must share a common definition of the situation. It is thus the spatially pre-defined situation that brings all actors into a conducive position, with their respective complementary social roles. The built environment thus delivers a necessary precondition of determinate social interaction.

For this to succeed the built environment must be legible. The participant can then respond to the spatial communication that is broadcast by the designed space, e.g. by entering a space and joining the accommodated social situation. As a communicative frame, a designed space is itself a communication as premise for all communications that take place within its boundaries.

The designed spaces deliver the necessary pre-definition of the respectively designated social situation, thereby reducing the otherwise unmanageable excess of action possibilities that exist in our complex contemporary societies. They ‘frame’ social interaction. The organistion and articulation of these framing spatial communications is architecture’s core competency. The social meaning of a space can usually be guessed from its location, shape and stylistic markers. The research programme of architectural semiology aims to analyse the active semiological codes that already operate within the built environment via spontaneous semiosis. There is also a design ambition to upgrade the communicative power of the built environment, project by project, via the design of information-rich systems of signification that aid navigation via way finding systems and aid interaction via the differentiation and nuanced spatio-visual characterisation of interaction offerings.

The success of such an endeavour obviously depends on the user uptake. This can probably expected in large, complex integrated social environments, as in a university or creative industry corporate campus, where life is communication intensive, where orientation is non-trivial, and where inter-awareness, knowledge transfer and ramifying collaborations put a premium on orientation and social participation. The question arises how the communicative performance of large, complex designed environments might be evaluated. The research project ‘Agent-based Parametric Semiology’ builds, investigates and applies a new form of crowd or occupancy modelling as an answer to this question.

 

2. AGENT-BASED OCCUPANCY SIMULATIONS

While every architect has an intuitive grasp of the normative ineraction protocols that attach to the various designated areas that the design brief indicates and usually knows enough about the expected and desired user occupancy patterns, such intuitions cannot give a secure guidance on the relative social performance of alternative designs for large, complex corporate environments. Intuition must here be substituted by simulations that can process thousands of agents interacting across an environment of hundreds of spaces. When quantitative comparisons and optimization is aimed at, then intuition fails already in much smaller, simpler settings.

The simulation methodology developed under the research agenda ‘Agent-based Parametric Semiology’  is conceived as a generalisation and corresponding upgrade of the kind of crowd simulations currently offered by traffic and engineering consultants concerned with evacuation or circulation.  These crowd modellers treat users as physical bodies and simulate crowds like a physical fluid. In contrast, architectural design considerations are concerned with socialized actors who orient and interact within a semantically differentiated environment. The simulations that must be developed to get a handle on desired social interaction scenarios will have to be different and rather more elaborate. They contain circulation models as trivial component.

The first and most obvious difference is the expansion of the menu of action types. The second major difference is that the agent population should be socially differentiated rather than homogenous.  For instance within the domain of corporate office life, agent differentiation might track rank, functional role, and team affiliation. An intricate network structure might be read off the client’s intra-net to inform the agent population within the simulation. The third significant difference and upgrade is the designation dependency of the agents’ behaviours. The designed environment is zoned in the sense of specifically designated and semantically encoded areas. Agents change their interaction propensity accordingly. This ordering increases the probability of highly specific interactions.  Where the number of designations and protocols to be distinguished is very large, it is opportune to use the combinatorial power of grammar to articulate this manifold. This thus implies a fourth enhancement, namely the elaboration of an agent system with semiologically competent agents.

The fifth aspect that distinguishes these architectural-semiological models from the circulatory crowd models is the following: Congenial with contemporary cultural conditions the underlying presumption of these models is that agents are largely self-directed, rather than running on pre-scheduled tracks, and do self-select their interactions and the social events they participate in. These selections are guided by multi-dimensional, dynamic utility functions that can utilise contingent opportunities that the agents encountered within the environment they browse through. These utility functions are implemented in the decision processes that control the agents’ actions on the basis of internal states due to prior actions and environmental offerings perceived. Within the preferred societal domain investigated by this research project – the domain of corporate working life -  the increasingly widespread use of non-territorialised, activity based office landscapes is most congenial to the assumptions and capacities of the life-process simulations promoted here. The sixth significant difference is that the focus shifts from the aggregation of parallel individual actions to the simulation of social interactions. This is most expressed in the case of corporate working spaces, as well as in university life, or in the case of conferences. In each of these arenas the purpose of gathering actors is precisely the opportunity of encounters and communicative exchanges.

The seventh, and here final, important differentiating aspect of the new life-process simulations is the fact that there can be no single generic agent model that could be transferred from project to project. The methodology invokes so many variables that it makes sense to speak of tailor made models for each specific domain of life, for each type of institution, and then further customized for each specific client. Not only are the agent populations and typical action menus as well as interaction propensities varied, but the very aims and criteria of success might vary. The model should attempt, as much as possible, with respect to available information, to model the actual user group, its differentiation into status groups, roles, even its established network of relations as might be retrieved via questionnaires or various electronic communication records. This customisation of the agent model does not only serve to enhance the veracity of the simulation, but also the information richness of the simulation results and the related ability of the methodology to home in on particular success criteria. A corporate client’s change management agenda might involve encounter and interaction facilitation between particular status groups, roles or even the empowerment and accessibility of particular important individuals. Such agendas can be catered for by tailoring the success criteria of the simulation accordingly.

In summary, here is the list of innovations that the research group is working on and that must be delivered by a semiologically informed social life-process modelling:
1.   expansion of the behaviour and action repertoire
2.   differentiation of agent population
3.   designation dependency of behaviours
4.   information empowered, semantically competent agents
5.   agent decisions via dynamic utility functions
6.   focus on social interactions and event scenarios
7.   domain tailoring and client customization

All these differentiating features imply challenges and necessary complications, or positively phrased, necessary sophistications for this much more ambitious modelling and simulation effort.

 

3. COMPUTATIONAL IMPLEMENTATION STRATEGY

The research team is currently building up increasingly large, differentiated and sophisticated agent populations using Unity game development software as base system, augmented with original coding. The development work concerning the agent populations benefits from a technology transfer from the game development industry, both with respect to basics like action animation and simple tasks like pathfinding, obstacle avoidance etc., and with respect to the more complex decision making processes modelled in ‘game AI’. Sophisticated games populate their gaming scenes with increasingly versatile, intelligent, seemingly spontaneous life-like agents.

Corporate work-life, the primary arena of application envisaged, entails multiple agent types and many different action types, each dependent on multiple conditions. This implies the requirement for the agent system to handle a large number of input factors for each agent and to use these to select from a large number of possible actions at each moment in the ongoing life-process of the model. This situation is too complex to work with the simplest and most widespread game AI methodology of ‘Finite State Machines’ (FSM). Finite State Machines work by defining states and conditions for transitioning between states. They are very open-ended and flexible by being able to translate from any state to any other state by specifying conditions. However, this approach lacks an ordering principle that could help game developers to systematically work through the complexity of human decision making. Beyond a certain complexity threshold it becomes too difficult for developers to map out all scenarios. The remedy for this shortcoming is the methodology of Hierarchical Finite State Machines reducing complexity by separating certain behaviours into sub-states. This lead to the idea of organising tasks in tree-like structures, so called ‘Behaviour Trees’, currently the dominant technique used by the game development industry. This methodology orders all actions into a hierarchy with a (left-to-right) priority sequence at each level. The decision of which action to execute is evaluated at a certain tick-rate.  At each tick the whole tree is walked through until conditions are met that trigger action. Complex behaviour trees can allow for concurrent actions as well as for sequences of behaviours. For large behaviour trees, the computational costs of evaluating the whole tree become prohibitive. Sub trees can then be introduced so that a sub tree can continue executing without invoking the whole tree, until some exit condition is met. However, this brings back the complexities of Finite State Machines.

The latest game AI methodology that is becoming more widespread in the game development industry is a methodology employing utility functions, so called ‘Utility AI’. Instead of switching between a set of finite states based on conditioning via triggers, or moving through a whole decision tree until trigger conditions are met, in Utility AI agents constantly assess the actions available to them in their current environment, and assign a utility score to each of those actions on a continuous scale. The utility system then selects the behaviour option that scores highest amongst the currently available options, based on the circumstances. Circumstances are both external and internal states. The latter being dependent on what went on in the game or simulation so far, i.e. the current utility and thus the urgency of a desire, and thus the utility of the related action, recedes or drops after the action was successfully completed and the desire satisfied. The basic laws of subjective economics like the law of diminishing marginal utility can be thus be implemented here. The normalized utility functions bring the most divers and otherwise incommensurable measures into direct comparison. Each choice of action is relative rather than based on absolute conditionals. These are temporary prioritizing decisions, based on internal states like desires, their urgency, available energy levels, as well as on opportunities afforded by environmental offering in proximity to current location. There can still be absolute conditionals placed in front, i.e. the various designated zones pre-condition the available action menu.  Utility AI can take any group of action options, destination objects, interaction chances and score these. This makes the methodology very versatile for decision-making. This approach comes closer to model emergent behaviour and realistic decision-making under incomplete information.

This technology transfer from the gaming industry delivers thinking tools, formalisation strategies and coding techniques for the elaboration of sophisticated autonomous agents capable of navigation and interaction within semantically charged environments. However, the specification of the appropriate action types and of realistic decision rules for the domain of life to be simulated had to be elaborated without readily available sources of knowledge. The members of our development team have to reflectively retrieve and rely on their own knowledge and experience of office life and working environments. Thankfully the domain of study  - creative industry work environments -  are a domain we all know from our own daily life experience. The task thus is to make our intuitive competency as well-socialized users of office environments and participants in corporate life explicit and then formalize those into behavioural dispositions and protocols. The same kind of reflection is required to tease out plausible desires, goals and quantified utility functions. Critics might question the veracity of these constructions and worry about problematic degrees of subjectivity. However, some of these competencies are obvious enough and we are here in a similar position to linguists, or more specifically grammarians, who use their own language competency as a basis and testing bed for the formalisation of the rules of grammar. Similarly, theoretical micro-economics is based on a priori assumptions about utility maximizing choices under constraints that are corroborated by reflection on our own economic choice competency rather than via empirical data. This does not mean that empirical observation should be excluded. It is indeed one of the ambitions and components of the research project to include empirical data collection as a means for model calibration. However, the efforts that must go into this are considerable, and the research and tool development should not have to wait for this.

The agent populations developed are being released into systematically varied designed environments. The resultant interaction events are then recorded to evaluate these environments with respect to defined criteria of success. The research focusses on communicative interaction events. These are the events we record, classified according to the duration, the number of participants, and the types of participants, i.e. their rank, and       whether the interaction was an inter- or intra-team event. The final success criteria that determine the evaluation of alternative office space designs will be specified in accordance to the client’s particular management agendas. Once success criteria have been formulated and can be measured, they can be aggregated into an overall fitness measure that in turn can be used to set up an evolutionary optimisation loop. The methodology can thus be made generative. The research team is in the process of making this move. Since the simulation effort is non-trivial, acceleration mechanisms and viable shortcuts are being explored and developed in parallel. Also, the computational burden of the simulations demands that design variations are strategically guided rather than random.

The question of the realism of the simulations is a difficult one. Empirical experimentation and calibration will eventually be required to enhance confidence here. That will be the role of the ZHA test bed where observations and sensor data collections are under way. However, we should also realize that absolute success measurements are not necessary to achieve a comparative evaluation and selection of the best design alternative. The presumption that relative performance advantages have veracity even if absolute performance measures are false, seems plausible enough to merit investing in the simulation methodology, especially if the alternative is the designer’s overburdened intuition, if not ignorance. Finally, the advantage of the Unity based simulations is that they also deliver immersive visualisations that can be used as intuitive plausibility and desirability check by both designers and clients.

 

4. SIGNIFICANCE FOR ARCHITECTURE

Architecture is one of the oldest academic disciplines and theory-led professions, and yet it remains one of the least developed in terms of the clarification of its core competency and in terms of its scientific underpinnings. The traditional unity of architecture and building engineering has been gradually dissolved and replaced by a thoroughgoing division of labour. The building engineering disciplines have been benefitting from a rigorous application of the hard sciences and their methodology. Architecture was left behind in a craft-like state, relying on accumulated experiences and tacit knowledge being carried forward from generation to generation via apprenticeships without receiving or benefitting from systematic explicit formulation. Architecture is presumed to deal with the ineffable, with matters that resist scientific treatment. Architecture is understood to deal with supposedly ineffable notions like order, aesthetics and meaning, perceived to be largely aloof from the mundane functioning of buildings.   

It his high time to discard such mystifications and start the process of turning architecture into a modern, science-based practice.

Architectural design, is constraint by engineering concerns, but should positively be driven by the social functions buildings are meant to perform. It is the categories, concerns and methodologies of the social and human sciences, not of the natural sciences, that need to be drawn upon here. The efforts of pioneers like Christopher Alexander and Bill Hillier have been recognized and admired but ultimately remained voices in the wilderness without impacting on the discipline’s ways of thinking and working. This also applies to the 1970s design methodology movement as a whole. What it delivered was obviously not compelling. The same is true for the 1970s efforts to establish a viable architectural semiology.

The design research programme of Agent-based Parametric Semiology picks up these loose ends and attempts to do better with a new, theoretically clarified and computationally empowered design methodology. This methodology brings the social function and meaning of architecture into the design model itself, in the form of semantically guided agent interactions, thereby operationalising and thereby upgrading architecture’s core competency of shaping the built environment to facilitate social communication and cooperation processes.

 

References

Schumacher, Patrik, Parametric Semiology – The Design of Information Rich Environments, in: Lorenzo-Eiroa, Pablo & Sprecher, Aaron (eds), Architecture In Formation, Taylor and Francis, New York, 2013